The integration of the geometrical ALS and radiometrical VHRS information made it possible to achieve high
classification accuracy 0.86, Figure 2 and generate segments of vegetation with a similar: height and spectral response. The
OBIA processing was performed in eCognition Developer 8.7 Trimble Geospatial and created rule-set was very universal
and not dependant on any samples. The details of the classification were a subject of different paper Tompalski and
Wężyk, 2011. The result of the OBIA processing, stored as a vector layer
Shapefile containing: buildings, low and high vegetation polygons, was subsequently used for batch processing of the
ALS point cloud inside each segment. The process was facilitated and automated with the use of LAStools LAStools,
2011 and FUSION
USDA Forest Service software McGaughey, 2010. Finally, a script written in R software
statistical package was applied to convert ALS point clouds into voxels 3D matrix and to calculate all the attributes needed
for computing 3D spatial indices. The volume of each building was calculated as the
multiplication of building footprint area and its height. The volume of each homogenous high vegetation class segment was
determined using voxels.
Figure 2. OBIA classification result – vector layer representing
land cover classes as a base for calculating the spatial indices. It has to be pointed out, that the above presented method of
extracting building, high and low vegetation areas, is probably one of many possibilities of achieving the result. The essential
subject of presented paper is not the method of classifying mentioned land cover classes, nor the accuracy of determining
the volume of elevated vegetation, but the spatial indices that rely on them. The input vector data can originate from any GIS
source, however it has to have the crucial attributes
– cubic volume of the high vegetation areas and buildings.
3. 3D SPATIAL INDICES
The general idea of the 3D spatial indices is to present diversity of different city regions concerning the relation between
vegetation and buildings. Two main 3D spatial indices are proposed:
1. Vegetation Volume to Built-up Volume VV2BV,
and 2.
Urban Vegetation Index UVI. The VV2BV is a ratio of the vegetation cubic volume to the
built-up volume and takes into account only the high vegetation class. It can be described with the formula:
1 where V
HV
is the total cubic volume of high vegetation and V
B
is the total cubature of built-up class. This index can be also
calculated as the percentage of high vegetation in the total volume of built-up and high vegetation. In this case the formula
is slightly modified:
2 The UVI index characterizes not only the high vegetation class
and built-up cubature, but also the area of low vegetation. It is calculated as a weighted sum of two ratios: high vegetation
volume V
HV
to built-up V
B
volume ratio and vegetation area to built-up area ratio. The relative area of height vegetation and
buildings is used as the weight w, Formula 4: 3
4 where A
V
is the total area of all vegetated surfaces low and high vegetation, A
HV
is the total area of high vegetation only and A
B
is the total area of built-up. As in the case of previous index, UVI can be as well calculated in relative values:
5 The use of relative area of elevated surface high vegetation and
buildings is used to split the formula into two parts. The first part is in fact the VV2BV index Formula 1 and is calculated to
express the relation between volume of high vegetation and buildings. The second part is meant to reflect the ratio between
the areas that are covered by all vegetation classes low and high to the area of buildings.
The proposed 3D spatial indices, in both cases have two types, depending on the object they are calculated for. The first type
VV2BV
CELL
or UVI
CELL
is calculated for each grid cell of the study area. The user can specify the cell size e.g. 100
×100m, 1km
×1km. In the second type of the index VV2BV
BUILDING
or UVI
BUILDING
calculated value is assigned to each building polygon and is based on the GIS spatial analysis performed
within a certain distance buffer area around each building e.g. 100m, 1km.
XXII ISPRS Congress, 25 August – 01 September 2012, Melbourne, Australia
174
Figure 3. 3D spatial indices calculated for the test area. Top images: VV2BV and UVI indices calculated for 100x100m cell grid, all test area is presented. Bottom images: VV2BV and UVI indices calculated for each building with buffer zone of 100m. Only a subset
of sample area is presented to preserve the details. In total, each of the two presented indices has four variants,
depending on the object they are calculated for cellbuilding and whether they have an absolute or relative value.
The method used to calculate each index can be explained on an example for single cell of 100
×100m, where buildings, low and high vegetation cover given area. The simplest example worth
considering is when all classes cover 13 of the cell area, and buildings and trees have both the same cubic volume. In this
case the indices would have values equal to: UVI
CELLabs
=1.33, VV2BV
CELLabs
=1, UVI
CELL
=55.56, VV2BV
CELL
=50. Together with the increase of the building area or volume, the
indices values would decrease. Since the volume is taken into consideration, similar index value can be achieved for a cell
with many small buildings or with single high building.
4. RESULTS AND DISCUSSION
